Articles & Publications

Artificial Intelligence & Architecture

From Research to Practice

Book | Birkhauser, '22

Artificial Intelligence’s (AI) encounter with Architecture is still in its infancy. However, current experiments and applications already are a testimony to their gradual intersection.

This book provides an introduction to the topic through the triple lens of History, Application, and Theory. A chronology of Architecture’s technological evolution first puts AI back in the context of the discipline. The author then presents a collection of AI’s applications in Architecture. The book finally gives the stage to contributors working at the forefront of this revolution. From Harvard to Foster & Partners, their perspectives provide a panorama of the discourse surrounding AI’s presence in the field.

Halfway between research and practice, this book offers to unveil the promise and challenges AI holds for Architecture.

With the contribution of Foster + Partners' ARD Group, the City Intelligence Lab, Kyle Steinfeld, Andrew Witt, Alexandra Carlson & Matias del Campo, Caitlin Mueller & Renaud Danhaive, Immanuel Koh, and Carl Christensen.

  • Publisher: Birkhauser
  • Published: 7 Mar. '22
  • Pages: 208
  • Language: English
  • on Birkhauser's website ->
  • on Amazon's website ->


Towards a Semantic Age for Architecture

Article | in "Artificial Intelligence & Architecture", Birkhauser, '22

For a long time, Architecture has benefited from fruitful analogies with linguistics. The language is a rich matrix that provides both a system and a free canvas where creations are expressed according to rules and transgressions. Quite naturally, architects have harvested its lexicon and frameworks to describe and think about Architecture.

Over the past decades, the discipline has in fact considerably borrowed from grammar and its concepts: the translation of Architecture into formal languages has corresponded to a need to formulate, organize, and replicate architectural information. Although this effort has proven to be very instructive for the discipline, a strict grammatical conversion does not fully account for many aspects of Architecture: at the very least it represents a missed opportunity.

Today, we believe that semantics offers a new angle to revive the analogy between Architecture and linguistics. This alternate approach should allow for a more adequate dialogue between technology and the architectural agenda. Built upon the latest development in Artificial Intelligence, we will call “Semanticism” this new momentum for Architecture.

Article published in "Artificial Intelligence and Architecture, From Research to Practice", Birkhauser, April '22

AI & Architecture

Towards a New Approach

Thesis | Harvard University, '19

The advent of Artificial Intelligence as a discipline has been permeating countless fields, bringing means and methods to previously unresolved challenges, across industries. The fusion of this new techno-science with Architecture is still in its early days, but offers promising results. Our thesis proposes to evidence its potential as it is applied to Architecture. More than a mere opportunity, it is, to us, shaping a promising discipline per se. Specifically, we offer to apply AI to floor plans analysis, and generation. Our ultimate goal is twofold: formulate a proper classification methodology of floor plans, able to tackle diversity and quantity, while creating a framework for machine learning-based floor plan generation.

Floor plans are indeed a high-dimensional problem, at the crossroads of quantifiable techniques, and more qualitative properties. The study of architectural precedent remains too often a hazardous process that negates the richness of the number of existing resources, while lacking in analytical rigor. We offer here a methodology, inspired by current Data Science methodologies, to qualify plans, both through their style and their organization.

At the heart of this project, lies the necessity of inventing meaningful metrics to qualify and classify floor plans. Through the creation of 6 metrics, we propose a framework that captures architecturally relevant parameters of floor plans.

On one hand, Footprint Shape, Orientation, Thickness & Texture are 3 metrics capturing the essence of a given floor plan’s style. On the other hand, Program, Connectivity and Circulation will capture the essence of a given floor plan organization.

The advent of automated classification is in fact the bedrock of most machine learning practices. Our thesis offers to leverage our database of classified plans to evidence the possibility of such tools applied to floor plan generation. Our methodology follows two main intuitions (1) the creation of floor plans is a non-trivial technical challenge that encompasses standard optimization techniques. (2) The design of space is a sequential process, requiring successive design steps across different scales (urban scale, building scale, unit scale). We attempt to capture these two realities by using nested Generative Adversarial Neural Networks. The use of such models will enable us to capture more complexity across encountered floor plans, while breaking down the complexity of the tackled problems into successive steps. Each step will correspond to a given model, trained for this particular task.

Overall, our thesis attempts to evidence the possible back and forth between humans and machines, that permeates the architectural discipline today. The machine, that was once the extension of our pencil, can today be leveraged to map architectural knowledge, and be trained to assist us creating viable design options.

Article published in "Artificial Intelligence and Architecture, From Research to Practice", Birkhauser, April '22

  • on Towards Data Science ->
  • on Archinect ->
  • on ArchDaily ->

Architecture as a Graph

A Computational Approach

Article | with Jeffrey Landes, Hakon Fure & Hakon Dissen

The design of floor plans can leverage machine intuition to generate and qualify potential design options. In this article, we address a specific abstraction of space: adjacency. Any floorplan carries its own embedded logic; in clear, the relative placement of rooms and their connections is driven by a certain logic of interdependence, and yields varying qualities across space. For instance, the presence of a room will condition the existence of other rooms, as well as the position of openings between them. First, we attempt here to qualify adjacencies of existing floor plans, to assess the relevance of adjacencies among rooms. We later turn to Bayesian modeling to generate adjacency graphs, either freely or under set constraints. By qualifying and generating, our hope is to investigate both sides of the same problem: the understanding of relationships among neighboring spaces.
  • on Towards Data Science ->

Space Layouts & GANs

GAN-enabled Floor Plan Generation

Article | Towards Data Science

Apartment layout is a challenging yet fundamental task for any architect. Knowing how to place rooms, decide their size, find the relevant adjacencies among them, while defining relevant typologies are key concerns that any architect takes into account while designing floor plans. In this article, we propose showcasing possibilities offered by Generative Adversarial Neural Networks models (GANs), and their ability to generate relevant floor plan designs. In short, we turn to GAN models, and more specifically Pix2Pix, to help us design housing floor plans, given a set of initial conditions & constraints.
  • on Towards Data Science ->

Architecture & Style

A New Frontier for AI in Architecture

Article | Towards Data Science

We build here upon a previous piece, where our emphasis revolved around the strict organization of floor plans and their generation, using Artificial intelligence, and more specifically Generative Adversarial Neural Networks (GANs). As we refine our ability to generate floor plans, we raise the question of the bias intrinsic to our models and offer here to extend our study beyond the simple imperative of organization. We investigate architectural style learning, by training and tuning an array of models on specific styles: Baroque, Row House, Victorian Suburban House, & Manhattan Unit. Beyond the simple gimmick of each style, our study reveals the deeper meaning of stylistic: more than its mere cultural significance, style carries a fundamental set of functional rules that defines a clear mechanic of space and controls the internal organization of the plan. In this new article,we will try to evidence the profound impact of architectural style on the composition of floor plans.
  • on Towards Data Science ->

Suggestive Computer-Aided Design

Assisting Design Through Machine Learning

Article | In collaboration with Thomas Trinelle

The utilization of machine-based recommendation has been leveraged in countless industries, from suggestive search on the web, to photo stock image recommendation. At its core, a recommendation engine can query relevant information -text, images, etc- among vast databases and surface it to the user, as he/she interacts with a given interface. As large 3D data warehouses are being aggregated today, Architecture & Design could benefit from similar practices.

In fact, the design process in our discipline happens mostly through the medium of 3D software (Rhinoceros 3D, Maya, 3DSmax, AutoCAD). Might it be through CAD software(Computer-Aided Design), or today BIM engines (Building Information Modeling), Architects constantly translate their intention into lines and surfaces in 3D space. Suggesting relevant 3D objects, taken from external data sources, could be a way to enhance their design process.

This is the goal of this article: study and propose a way to assist designers, through “suggestive modeling”. As architects draw in 3D space, an array of machine-learning-based classifiers would be able to search for relevant suggestions and propose alternative, similar or complementary design options.

To that end, taking inspiration from precedents in the field of 3D shape recognition & classification, we come up with a methodology and a toolset able to suggest models to designers as they draw. Our goal is, in fact, twofold: (1) to speed up the 3D-modeling process with pre-modeled suggestions, while (2) inspiring designers through alternative or complementary design options.

  • on Towards Data Science ->

The Advent of Architectural AI

A Historical Perspective

Article | Towards Data Science

The practice of Architecture, its methods, traditions and know-how are today at the center of passionate debates. Challenged by outsiders, arriving with new practices, and questioned from within, as practitioners doubt its current state, Architecture is undergoing a truly profound (r)evolution.

Among the factors that will leave a lasting impact on our discipline, technology certainly is one of the main vectors at play. The inception of technological solutions at every step of the value chain has already significantly transformed Architecture. The conception of buildings has in fact already started a slow transformation: first by leveraging new construction techniques, then by developing adequate softwares, and eventually today by introducing statistical computing capabilities (including Data Science & AI). Rather than a disruption, we want to see here a continuity that led Architecture through successive evolutions until today. Modularity, Computational Design, Parametricism and finally Artificial Intelligence are to us the four intricate steps of a slow-paced transition. Beyond the historical background, we posit that this evolution is the wireframe of a radical improvement in architectural conception.

  • on Towards Data Science ->

K.I. und Architektur

Der entwerfende Computer

Article | in "Posthumane Architektur", ARCH +

Wie lassen sich Computer, die früher lediglich eine Art erweiterter Zeichenstift waren, heute dazu nutzen, architektonisches Wissen abzubilden? Wie lassen sie sich mit Hilfe maschinellen Lernens trainieren, Architekt*innen bei der Entwurfsarbeit zu unterstützen?

Die hier vorgeschlagene Methode basiert auf einem dreistufigen Prozess: (I) Generierung von Grundrissen, das heißt Optimierung der Erzeugung sehr unterschiedlicher Grundrissentwürfe in großer Zahl, (II) Qualifizierung von Grundrissen, das heißt Entwicklung einer geeigneten Klassifizierungsmethode, und (III) Bereitstellung der Möglichkeit für die Nutzer*innen, die erzeugten Entwürfe zu durchsuchen.

Der Prozess zielt auf eine bessere Differenzier-, Qualifizier- und Modulierbarkeit von Ergebnissen im Vergleich zu bisherigen Programm­anwendungen aus dem Feld der KI-gestützten Raumplanung.

  • on ARCH + ->

Urban Tech On The Rise

When Machine Learning Disrupts the Real Estate Industry

Article | in FACTS Reports, with Daniel Fink & Pamella Gonçalves

The practice of urban analytic is taking off in the real estate profession. Data science and algorithmic logic are close to the forefront of new urban development practices. “How close?” is the question, but experts consider that digitization will go far beyond intelligent building management systems. New analytical tools with predictive capabilities will dramatically affect the future of urban development, reshaping the real estate industry in the process.
  • in the Harvard Real Estate Review ->
  • in the Veolia Institute FACTS Report ->